NOTES
1.
2.
3.
4.
5.
6.
7.
See Lee and Wang (1999) for an extended discussion of Malthus' continuing influence on analyses of Chinese population and society.
The positive check also played a role, but not in the way that Malthus thought.
Whereas Malthus and others emphasized the adverse effects of misery and poverty on death rates across the age range, Lee and Wang argue that the calculated use of infanticide by couples to adjust family size and sex composition to their circumstances predominated.As discussed in Bengtsson, Campbell, Lee et al. (2004, ch. 2), composite indices of standard of living constructed for countries by the United Nations, the World Bank, and other organizations almost always include a measurement of the level of mortality, often the infant mortality rate but sometimes life expectancy.
Though historical studies of long-term trends in mortality in China based on lineage genealogies such as Liu (1985) and Harrell (1995a) have remarkable time depth, in some cases extending back into the Ming dynasty (1368—1644), the limitations of retrospectively compiled lineage genealogies as a source have also prevented such studies from examining death rates in the late nineteenth century and early twentieth century.
See Appendix A of Lee and Campbell (1997) for a detailed introduction to these data.
According to the traditional Chinese procedure for calculating age, individuals were 1 sui at birth, and aged by 1 sui every Lunar New Year. As a result, ages reckoned in sui are on average 1.5 years higher than western ages.
Unfortunately none of the registers from between 1888 and 1903 has been located, thus a precise accounting of this increase is not yet possible.
Note that even though Daoyi is in the central region, and very close to Guosantun and Daxingtun, we treat it separately here since we have studied it previously and already have empirical results for it.
See Lee and Campbell (1997: 66—8) for an adjustment of fertility estimates for Daoyi that made use of indirect methods. We concluded that multiplying the TMFR based on surviving male births by 2.91 yields a reasonable estimate of the level of fertility based on all births. This adjustment is the product of two factors: 1.5 to account for boys who died without ever being registered, and 206/106 to account for female births.
10 Another limitation of the registers is that they do not allow married women to be traced back to their natal families; thus we do not know whether families obtained brides for their sons from within the same register population, or from other populations in the area. In other words, we cannot study intermarriage.
11 See Lee and Campbell (1997: 31—5) for additional details on the grain price series.
12 Many of the most prominent spikes were associated with extended spells of unfavourable weather such as cold summers (Lee and Campbell 1997: 34).
13' We also included main effects of region, operationalized as a set of three dichotomous indicator variables for central, Daoyi, and south. North was the omitted category. To save space, we do not present the coefficients for these terms.
14 Complementary log—log regression is more appropriate for event history than logistic regression because it yields coefficients that are comparable to those from continuous-time proportional hazards models (Long 1997).
15' Because marriage rates varied substantially by age within each interval, we included a fourth-order orthogonal polynomial in age, along with interactions between the terms of the polynomial and region. To save space, we do not present the coefficients for these terms.
6 To account for variation in age-specific probabilities of dying within these age ranges, we include a quadratic
orthogonal for age, interacted with the region indicators. Once again, to save space we do not present the
coefficients for these terms.
17' To save space, we do not present the estimated coefficients here.
18' To save space, we do not present the estimated coefficients here.